Geekbench has introduced a cross-platform tool designed to evaluate device performance under AI-intensive tasks. Named Geekbench AI, it evaluates how well a device’s CPU, GPU, and NPU handle machine learning applications. By focusing on AI-heavy workloads, the tool provides insights into how different hardware configurations perform in the realm of AI
Geekbench AI assesses performance based on both speed and accuracy, supporting various AI frameworks such as ONNX, CoreML, TensorFlow Lite, and OpenVINO. This would take a comprehensive evaluation of how effectively different devices manage different types of AI-based tasks.
The tool generates three distinct scores for its evaluations: full precision, half precision, and quantised. It includes an accuracy measurement to determine how closely a device’s outputs match expected results. This aspect is important for understanding how accurately a device performs AI-related functions, such as machine learning model predictions.
As it becomes available for download on multiple platforms, including Windows, macOS, Linux, Android and iOS, we will see how real-world AI performance aligns with Geekbench AI’s metrics. Future assessments might extend beyond traditional benchmarks to include evaluations of predictive text accuracy and the quality of generative AI applications.
